Based on grey system forecasting model, neural network forecasting model and support vector machine forecasting model, the linear combining forecasting model, neural network combining forecasting model and combining support vector machine forecasting model are set up. Taking the gross of waste water data of Changjiang from 1995 to 2004 as a study case, compared with single prediction methods, linear prediction method and neural network combining forecasting method, the accuracy of the support vector machine method is higher.